memory-tree
Pass
Audited by Gen Agent Trust Hub on Apr 7, 2026
Risk Level: SAFECOMMAND_EXECUTIONPROMPT_INJECTION
Full Analysis
- [PROMPT_INJECTION]: Indirect Prompt Injection Surface. The skill is designed to ingest, store, and retrieve arbitrary user-provided data, which can influence agent behavior in future sessions.\n
- Ingestion points: Untrusted data enters the system as the first argument to
scripts/memory-tree/curate.sh.\n - Boundary markers: No delimiters or instructions to ignore embedded commands are specified for the stored memory content.\n
- Capability inventory: The skill can execute shell scripts via
bash, perform file system writes in thememory/tree/directory, and interact with theqmdCLI tool.\n - Sanitization: The provided documentation does not describe any validation or sanitization of the input data.\n- [COMMAND_EXECUTION]: The skill documentation indicates the execution of several shell scripts (e.g.,
curate.sh,query.sh,migrate.sh) that accept user-provided strings as positional arguments. Without internal review of these scripts, there is a risk of command injection if the arguments are not properly escaped before being processed by the shell.
Audit Metadata